WO1992005478A1 - Control system - Google Patents
Control system Download PDFInfo
- Publication number
- WO1992005478A1 WO1992005478A1 PCT/CA1991/000338 CA9100338W WO9205478A1 WO 1992005478 A1 WO1992005478 A1 WO 1992005478A1 CA 9100338 W CA9100338 W CA 9100338W WO 9205478 A1 WO9205478 A1 WO 9205478A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- trip
- signal
- input signals
- active state
- activity
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
- G05B13/027—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks only
Definitions
- This invention relates to a control system and a method for operating a control system, and in particular a control system in which the external situation after actions of the system automatically result in the modifying of the strengths of certain signals internal to the system, thereby increasing the effectiveness of the system.
- a system for initiating an action in response to a plurality of variably weighted inputs has a layered configuration, and in the simplest arrangement comprises two layers.
- the first layer has:
- the second layer has: (a) means for receiving a second input signal resulting as a consequence of the first active state; and
- the system further includes means for timing the delay between the production of the trip signal and the suppression of the first active state and means for modifying the weighting applied to the first input signals as a function of the delay.
- the first input signals may be received from a first level of sensors and the second input signal received from a second level sensor.
- the first level sensors may be used to make an initial identification of a condition which is then confirmed by the second sensors.
- the trip signal is produced to initiate the first active state.
- the second active state is then initiated if the second sensor confirms the presence of the condition.
- the system is adapted to increase the weighting which is applied to the first input signals which contributed to the reaching of the predetermined threshold level of the first parameter.
- the weighting may be increased up to a maximum level. The increase in weighting in this situation occurs due to the short time interval between the production of the trip signal and the suppression of the first active state.
- the weighting applied to the first input signals which contributed to the production of the trip signal is decreased to minimise the possibility of the incorrect activation occurring again.
- the first active state continues unsuppressed as the second active state is not initiated. This continues for a predetermined time until an over-ride operates and suppresses the first active state.
- the time delay between the production of the trip signal and the suppression and the first active state will be relatively long.
- the delay timing means includes a lag function to take account of this time interval.
- the control system may have two or more layers, the only requirement being that one state is capable of definitively recognizing the desired condition.
- a system for initiating an action in response to a plurality of variably weighted inputs has a layered configuration, and may comprise two layers.
- the first layer has: a) means for receiving a plurality of first input signals and remaining active while receiving the first input signals; b) means for applying a weighting to the first input signals; c) trip means for receiving the weighted first input signals, summing the signals and becoming active when the sum of the weighted first input signals reaches a predetermined threshold level; d) means for initiating a first active state; and e) means for providing a positive feedback from the first active state means to the trip means to maintain the trip means active even if the first input signals receiving means becomes inactive.
- the second layer has: a) means for receiving a second input signal resulting from a consequence of the first active state; and b) means for initiating a second active state when the second input signal reaches a predetermined level. Initiation of the second active state suppresses the first active state and the positive feedback from the first active state means to the trip means.
- the system further includes means for measuring the temporal relationship between the activity of the first input signal means and the activity of the trip means, and means for modifying the weighting applied to the first input signals as a function of this temporal relationship.
- FIG. 1 is a block diagram of a control system in accordance with the preferred aspect of the present invention, as embodied in a plastic scrap labelling machine;
- Figure 2a-d are trip unit output graphs of connection strength against time
- Figure 3 is a more detailed block diagram of area 2 of Figure 1;
- Figures 4a-b are plots of the forcing function which controls the smoothing unit for strength change command;
- FIG. 5 is a block diagram of a control system in accordance with a further aspect of the present invention, as embodied in a collection machine;
- FIGS. 6, 7 and 8 are block diagrams of a control system in accordance with a still further aspect of the present invention, as embodied in a battery powered machine .
- the present invention relates to a control system and the method of operation thereof.
- the system will be described initially by way of example with reference to a specific application. The system will then be describee more generally and the operation of the system described in terms of mathematical functions. This is followed by a description of other specific applications to illustrate various features of the system.
- Figure 1 of the drawings illustrates a control system 10, in simplified block diagram form, forming part of a plastic scrap labelling machine for use in the separation of plastic scrap her material. The machine is intended to identify plastic scrap and label it ready for identification and collection by another machine.
- the control system is arranged in a plurality of layers, identified as 1, 2, 3, 4 and 5.
- Each layer includes one or more sensors S which respond to stimuli in the environment, a plurality of weighted connections W from the sensors S, a trip device R which receives and sums the weighted inputs from the connections W and produces a trip signal when the sum of a selected parameter of the weighted inputs reaches a predetermined threshold level, the trip signal causing a state unit B to become active and initiate a command for some form of action whose consequences, if appropriate, will activate the next layer of the system. On activation of the state unit B in the next layer, the state unit B in the previous layer is suppressed or deactivated.
- every trip device R receives signals across the connections W from each sensor S. These connections W have variable strength or weighting, and "learning” occurs through the weighting changes in the connections W. The weighting in the connections W between the sensors S and the trip device R are modified according to whether the action in the succeeding layer initiated by the previous layer results in the initiation of the successive active state unit B.
- the control system is formed of a number of layers, each carrying out a sub-task in successive steps.
- the machine controlled by the system moves over an area and when the first level of sensors S1 detects, for example, a scrap, which from a certain property of reflectivity or colour, appears to be plastic, the machine will "land" to take a closer look at the material using the second level of sensors S2 which detect surface texture. If the surface texture also suggests that the scrap is of a plastic material, the third level of sensors S3 are activated to chemically evaluate or "sniff" the plastic. If the sensors S3 confirm that the scrap is plastic, the machine will stick a label on the scrap and then leave, after the sensors S4 have ensured that the label is in the correct location on the scrap. There is an over-ride command detected by the sensors S5, to pick up the labels and leave if the machine has not successfully reached the initial leave stage (at level 4) within some preset time.
- Each set of connections W are initially weighted so that when the associated sensors S in a layer are presented with stimuli which the sensors have been set to identify as a feature of plastic, the level of activity in the sensors S and thus the level of input signals to the respective trip device R will increase.
- the learning modifications made to the connections W during the operation of the system strengthen the possibility that the trip device R will be pushed' over the predetermined trip threshold by "correct” objects and teach it not to respond to "incorrect” ones.
- plastic scrap is initially identified as being more reflective than other material that the machine is likely to encounter so that some of the sensors S1 respond to "shiny" objects. Responses to "wrong" shiny surfaces are eliminated through the learning process provided by the connections, as described below.
- the learning procedure is a trial and error process which can be thought of in terms of hypothesis formation and testing.
- the answer to the hypothesis test, provided by the next layer can be thought of as providing a teaching input which indicates whether the sensor identification was correct.
- the weighting of the connection W from a particular sensor S is altered such that the probability that particular stimulus pattern will result in activation of the trip device is appropriately altered.
- To "train" the connections W1 it is necessary to strengthen the response to real plastic scrap and weaken the response to patterns activated by objects which resemble plastic scrap but are not.
- connections W1 are weighted such that the trip device R1 is activated by an object with certain reflective qualities in the field of vision of the sensors
- the control system and machine has, in effect, hypothesized that the object is plastic scrap.
- the trip device RI activates the state unit B1.
- the state unit B1 commands the machine to land on the scrap.
- the visual pattern which activated the sensors S1 will disappear, but the feedback between the state unit B1 and the trip device R1 maintains activation of B1 and R1.
- the control of movement is subsumed by, for example, a contact reflex, and the descent stops. This does not interrupt activation of the state unit B1 or trip device R1, which continues due to the positive feedback loop F1 therebetween.
- After landing there will be a stable, visual image available again and the second level of sensors S2 produce signals based on the surface texture of the scrap.
- the closeup image input signals produced by the sensors S2 and weighted in the connections W2 will activate the trip device R2 which in turn activates the state unit B2 and the third layer of the system is activated and the machine commanded to "sniff" the object with the plastic chemo-detector S3.
- the state unit B2 has inhibitory connections (indicated by blunt-ended lines 12), which are set slightly stronger than any activating connection from the state unit B1 to state unit B2 (indicated by arrows 14). This ensures that state unit B2 shuts off or suppresses state unit B1.
- a trip device R1 now no longer has input from the sensors S1 nor positive feedback from the state unit B1 so that deactivating the state unit B1 will also deactivate trip device R1.
- the layer becomes inactive or silent.
- the learning process is also capable of refining the ability of the system to recognize plastic and this can be illustrated by the situation in which the sensors S1 include colour detecting sensors which are initially set with very weak or low-weighted connections W1 to the trip device R1.
- the colour detection sensors S1 which are activated by a particular piece of correctly identified plastic will have the connections W1 to the trip device R1 strengthened, because it is only the temporal relation of the activity patterns that controls the strength changes.
- the fact that the colour sensor connections W1 were initially so weak as to have negligible influence in pushing the trip device R1 above threshold is irrelevant. Following strengthening of the colour connectors W1, when a similar plastic object is seen again the colour will contribute a greater amount to the recognition process.
- the system allows already effective connections involved in performing successful identifications to self-reinforce, the increment caused by each reinforcement becoming smaller as the connection approaches maximum strength.
- the sensors S1 and connections W1 which identify highly reflective objects as plastic continue to do so, and the other sensors S1 and connections W1 gradually learn to identify objects on the basis of colour also, if there are reliable colour cues available.
- a system according to the present invention may also include more than two layers, as does the example described with reference to Figure 1, and permits information from an incorrect response to propagate backwards down the layers from, for example, the fourth layer of the system.
- a flaw in the label design results in a situation in which some forms of plastic scrap cannot be labelled by the machine.
- the machine will initially identify the plastic by way of reflectiveness and colour (layer one), confirm the surface texture of the plastic (layer two), identify and confirm by chemical detection that the material is plastic (layer three) and then attempt to label the scrap, which is the action commanded by the state unit B3.
- a "success" is when the sensors S4 detect a label correctly positioned on the scrap. This triggers the trip device R4 and activates the state unit B4 which, in turn, inhibits or turns off the state unit B3 which issued the command for labelling to take place, which in turn will result in the trip unit R3 being turned off.
- the sensors S4 will not detect the label in position and the trip device R4 will not be triggered and thus, the state unit B4 will not be activated to inhibit or turn off the B3/R3 complex. Therefore, the trip device R3 will not turn off until the over-ride activates, as detected by the sensors S5, long after the sensors S3 were active or the original trip signal was produced by the trip device R3.
- the pattern is sensor S3 "on” then "off", trip device R3 "on”. This is the connection weakening pattern shown in Figure 2d. Repeated weakening will lead to a situation where the machine simply will not try to label this form of scrap when it is detected by the sensors S3.
- the system will still lead the machine through the various identifying steps with the unlabelable scrap up to the second layer, where the state unit B2 activates operation of the chemical sensors S3.
- the state unit B2 will remain active until the over-ride operates.
- the failure to achieve a positive labelable plastic identification in the third layer will result in weakening of the connections W2 which lead from the sensors S2 which detect textural features of the unlabelable scrap.
- Figure 3 illustrates, in somewhat more detail, the components of the sensors S1 and the trip unit R1, and the mechanism whereby connection weight is altered. It should be noted that Figure 3 illustrates only a single sensor S1 to simplify the diagram and to facilitate understanding of the operation of the system.
- the sensor includes an input sensing unit 30 which supplies signals to a transmission or trigger unit 32.
- the trigger unit 32 is provided in the sensor so that inputs from the sensing unit 30 caused by noise and the like are not relayed as signals to the trip unit R1, signals only being passed to the trip unit R1 once the inputs reach a predetermined threshold level.
- the sensor S1 also includes an activity sensing unit 34, which records periods of activity of the sensing unit 30.
- Input signals from the trigger unit 32 pass through the connection W1, which includes a weighting unit 36, to an input sensing unit 38 of the trip unit.
- This input sensing unit 38 receives inputs from a plurality of sensors though, as mentioned above, only one sensor is illustrated in the Figure.
- a trigger unit 40 is activated and in turn activates B1 which provide feedback to keep the trigger unit 40 active, even if input from the sensors falls below the threshold level.
- An activity sensing unit 42 records the activity of the trigger unit 40.
- the weighting applied to the connectors W1 is varied depending on the temporal relationship between the activity or firing pattern of the sensor and the trip unit.
- the relative activity of the sensor S1 and the trip unit R1 is recorded in a relative activity sensing unit 44, linked to the activity sensing units 34, 42.
- the link between the activity sensing unit 34 of the sensor and the relative activity sensing unit 44 is subject to the time-lag ⁇ (tau) , the lag being effected by a lag unit 46.
- ⁇ time-lag
- each of the units 34, 42, 44 decay with time, but the value in unit 44 decays more slowly such that, for example, if the activity sensing units 34,
- the values held in the sensing units 34, 42, 44 decay over time. This is accomplished as follows.
- the values held in the units are modified by respective smoothing units. These are units that accept a signal and record a time-weighted average of the signal.
- ⁇ is called the decay parameter.
- connection W The mechanism for strength change at a variable strength S to R connection W is as follows.
- Each given connection has designed into it a time-delay, ⁇ provided by lag unit 46.
- ⁇ provided by lag unit 46.
- the connection strengthens or weakens according to the co-activity in R at time t and S at time t - ⁇ .
- ⁇ R the activity registers of the sensor S and trip unit R as held in the units 34 , 42 .
- These activity registers are smoothing units, as described above, recording a time- weighted average of the activities of S and R respectively.
- the signals received by ⁇ s or ⁇ R is 1 if the particular trip unit or sensor is on at time t, and 0 it is off at time t.
- the decay parameter has two possible values: ⁇ if the trip unit or sensor is on at time t, and ⁇ if the trip unit or sensor is off at time t, Here 0 ⁇ .
- d ⁇ /dt if the neuron is on at time t
- the strength change will depend on the "correlation" between ⁇ sd and ⁇ R .
- RASU the relative activity smoothing unit held or value in unit 44.
- A(t) is called the co-activity of S and R at time t, and can vary between in this example between -2 and +1.
- the co-activity then serves as the signal for the relative activity smoothing unit (RASU), whose value is held in unit 44.
- RASU relative activity smoothing unit
- ⁇ is a positive constant which is less than ⁇ , the decay constant for the activity registers.
- ⁇ ⁇ ⁇ is that the RASU remembers co-activity longer than ⁇ R and ⁇ s remember activity.
- connection weight register w(t)
- W max the connection weight register
- the effect of the factor (1- ⁇ sd (t)) MS (1- ⁇ R (t)) MR is that activity in R or S (which causes ⁇ s and ⁇ R to be close to 1) prevents w(t) from changing. That is, it inhibits consolidation.
- the magnitude of the rate of change of w(t) depends on the magnitude of the RASU register ⁇ (t).
- the sign of ⁇ (t) determines the direction of change of w(t), increasing towards W max if ⁇ (t) is positive and decreasing towards 0 if ⁇ (t) is negative.
- Figures 2a-d of the drawings illustrate a trip unit output graph of connection strength against time, depending on various possibilities of activity in the trip unit R and the sensor S.
- the parameters are set as shown in the Figure, and the manner in which the symbols above relate to the parameter names is shown.
- each 1 represents 10, or one TAUGAP.
- Figures 4a-b of the drawings plot the coactively function A(x,y) which is used to construct the signal to the relative activity smoothing unit RASU.
- the x axis is ⁇ s and the y axis is ⁇ R .
- the graph itself plots the surface on which A(t) moves while a. and ⁇ R vary.
- Path 1 shows that if a. was very low and ⁇ R very high, there is little tendency to change strength. Thus if S is off and R fires strongly, there will only be the smallest tendency to weaken the connection - virtually no change will occur if R fires alone briefly.
- Curve 2 shows that if ⁇ s is intermediate and ⁇ R is high when activity ends, the decay back to 0 will be mostly through the "trough" section (and of course the no change section, provided such that noise and 'blips' in the sensors S and trip units R will not influence weighting) and the connection will weaken. This is the situation in Figure 2d, where ⁇ s has been off long enough to have decayed considerably before R turns off. If the amount of connection weakening vs. the duration of R activity following S's switch-off is graphed, it is seen that there is a rather long range of R activity durations during which considerable weakening occurs.
- Curve 3 shows how proper co-termination will strengthen the connection, when switch-offs occur near the "correct interval of ⁇ time units apart, so that both ⁇ R and ⁇ sd are approximately the same.
- the path remains in the positive section for a long time, and cuts relativety directly across the trough, so net strengthening occurs.
- This curve also allows one to see graphically why it is important to set a high value for ⁇ , the constant determining how quickly ⁇ s and ⁇ R rise when activity begins. s too small, then after the usual bout of firing ⁇ s and ⁇ R will not be near 1. The path will begin to be followed part way down, say at (0.8,0.8), so the positive section of the path will not be as lorg, and strengthening will not dominate. Thus if ⁇ were much too small weakening would occur.
- the sensors Sa in the first layer identify a plastic container, a successful identification tripping the device Ra to activate a state unit Ba, which issues a command for the machine to collect the container.
- the sensors Sb detect the contents of the container and, if the contents are corrosive, trigger the trip device Rb which activates the state unit Bb to remove the container from the machine and simultaneously notify a central controller that corrosive material has been located.
- connections Wa between the sensors Sa and the trip device Ra are similar to those described with reference to the above-described labelling machine application in that the weighting or the strength of the connections Wa will be modified based on the temporal activity of the sensors Sa, trip device Ra and state unit Ba.
- the weighting of the connections Wa will be strengthened in circumstances where the state units Ba and trip device Ra is suppressed or turned off by the state unit Bbl. That is, when the collection action is immediately accomplished the connections Wa from the sensors Sa which detect containers will be strengthened. Some collected containers will contain corrosive material detected slightly later. In the situation where the second level sensors
- connections Wab between the sensors Sa and the trip device in the second layer Rb.
- These connections Wab are adapted to be strengthened in these circumstances.
- the connections Wab strengthen because the on-off activation of Sa is followed by the on-off activation of Rb2, and only co-activation patterns control strengthening.
- the strengthened connections Wab will trigger the trip device Rb to activate the state unit Bb to indicate a corrosive material has been located, without the intermediate step of placing the container in the machine.
- the system can thus learn on the basis of consequences beyond those immediately following B activation.
- the system may include time lag factors ⁇ (Tau) to accommodate time delays, in this case a longer time lag factor would be provided to accommodate the time interval between the plastic container being seen and being analyzed for corrosive contents.
- FIG. 6 illustrates a control system 60 for use in a machine which is electrically powered, power being provided by a re-chargeable battery.
- the system operates when the battery of the machine is on low charge and is intended to allow the machine to find a power outlet, to plug into the power outlet and thus, recharge the battery.
- the example is intended to illustrate a "deficit reduction" feature of the system.
- the control system is arranged in two layers.
- the first layer includes a trip unit Rd which is activated when the battery charge falls below a predetermined level, and continues firing while the battery charge is below this threshold level.
- the state unit Bd commands that the machine approach objects that the machine has visual sensors focused on that point in time, followed by an attempt to connect with a power outlet, and if no connection is made to move off in a different direction. A machine with a low battery will thus begin to approach various objects in the environment.
- connection Wd to the trip unit Rd are not subjected to strength changes, because the connection strength change requires that the trip unit Rd be silent or inactive during consolidation or strengthening (due to the inhibitory connections, such as those illustrated as 48 and 50 in Figure 3). Therefore, the connections Wd to Rd are not effected by unsuccessful actions, that is, by actions which do not precede recharging.
- FIG. 7 of the drawings illustrates a control system 100 for use in carrying out the same function as the system described with reference to Figure 5 of the drawings, but in which the control system is of a somewhat simpler nature.
- the first level sensor SX locates plastic containers, and as soon as the container is located the trip unit RX activates the state unit BX to collect the container and move on to locate the next container.
- the sensor SY detects corrosive material in the container which has been taken on board. On detecting corrosive material the trip unit RY is activated to produce a trip signal which activates the state unit BY which removes the container from the machine and notifies base.
- the unit BY also suppresses the unit BX (or any activity which takes place as a result of the container being collected).
- the connections WX, WY are "fixed" and they have no learning capacity.
- This connection is strengthened in situations where the sensors SY detect the presence of a corrosive material and allow the system to learn to identify containers which hold corrosive material without taking the container on board, as the strengthened connections WXY will identify plastic containers which carry corrosive material and will cause a trip signal to be produced by the trip unit RY to activate BY and thus suppress BX, preventing the container from being taken on board.
- the trip unit RX and the state unit BX include activity registers for maintaining activity values which are a function of the periods of activity of the respective trip and state units RX, BX.
- the system further comprises a co-activity register for maintaining a co-activity value which is a function of the activity values, the activity and co-activity being subject to respective decay functions, the modification of the weighting applied to the connection WXY being a function of the co-activity value when the activity of the trip unit RX and the state unit BX ceases.
- the activity registers include smoothing units and may accommodate appropriate lag times, as described above with reference to Figure 4 of the drawings .
- the control system 120 is part of a fire alarm system and includes a smoke detector DS and a high temperature detector DT. There is a connection WS between the smoke detector DS and a fire alarm BF.
- the smoke detector BS indicates where there is smoke in an area, and the high temperature detector where there is a high temperature in the area, as a result of fire.
- the connection WS between the smoke detector DS and fire alarm BF is insufficient to trigger the fire alarm, which initially can only be triggered by the high temperature sensor DT.
- the system may be "taught" to set off the fire alarm BF on detection of smoke by the smoke detector DS.
- the sensors must produce signals which stop after a certain period.
- the sensors are of the onset type, in which the sensors will create a burst of activity on sensing a certain condition.
- the weighting of the connection WS is a function of the time between the ends of the bursts of activity of the respective sensors DS, DT.
- the connection WS passes through a trip unit RS which determines the strength of signals from the sensor DS which are sufficient to activate the alarm BF.
- Both sensors DS, DT include activity registers AS, AT for maintaining activity values which are a function of the periods of activity of the respective sensors.
- a co-activity register AST for maintaining a co-activity value which is a function of the sensor activity values, the activity and co-activity being subject to respective delay functions and the modification of the weighting applied to the connection WS being a function of the co-activity value when the activity in both sensors DS, DT ceases.
- the measure of activity of the smoke detector DS is, in this example at least, required to include a time lag function ⁇ (Tau), as there will clearly be a delay between the detection of smoke and detection of high temperature resulting from fire.
- the activity and co-activity registers may operate in a similar manner as described above with reference to Figure 4.
- control system of the present invention may operate in a wide range of applications in addition to those described above. It will, of course, be clear to those skilled in the art that the above described embodiments are merely exemplary, and that various modifications and improvements may be made to the invention without departing from the scope of the invention.
Abstract
Description
Claims
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP3515417A JPH06501577A (en) | 1990-09-25 | 1991-09-23 | control system |
CA002091772A CA2091772C (en) | 1990-09-25 | 1991-09-23 | Control system |
DE69105145T DE69105145T2 (en) | 1990-09-25 | 1991-09-23 | CONTROL SYSTEM. |
EP91916369A EP0550501B1 (en) | 1990-09-25 | 1991-09-23 | Control system |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US07/587,913 US5191636A (en) | 1990-09-25 | 1990-09-25 | Control system with two level sensor system |
US587,913 | 1990-09-25 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1992005478A1 true WO1992005478A1 (en) | 1992-04-02 |
Family
ID=24351692
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CA1991/000338 WO1992005478A1 (en) | 1990-09-25 | 1991-09-23 | Control system |
Country Status (8)
Country | Link |
---|---|
US (1) | US5191636A (en) |
EP (1) | EP0550501B1 (en) |
JP (1) | JPH06501577A (en) |
AU (1) | AU8522291A (en) |
CA (1) | CA2091772C (en) |
DE (1) | DE69105145T2 (en) |
ES (1) | ES2063524T3 (en) |
WO (1) | WO1992005478A1 (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7769474B2 (en) * | 2005-09-20 | 2010-08-03 | Honeywell International Inc. | Method for soft-computing supervision of dynamical processes with multiple control objectives |
DE102008000038A1 (en) * | 2008-01-11 | 2009-07-16 | Robert Bosch Gmbh | contraption |
US8242899B2 (en) * | 2010-02-09 | 2012-08-14 | InnovAlaem Corporation | Supplemental alert generation device for retrofit applications |
US8237577B2 (en) * | 2010-02-09 | 2012-08-07 | Innovalarm Corporation | Supplemental alert generation device |
US8558708B2 (en) | 2010-02-09 | 2013-10-15 | Innovalarm Corporation | Supplemental alert generation device with speaker enclosure assembly |
WO2011100121A1 (en) * | 2010-02-09 | 2011-08-18 | Innovalarm Corporation | Supplemental alert generation device |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4873655A (en) * | 1987-08-21 | 1989-10-10 | Board Of Regents, The University Of Texas System | Sensor conditioning method and apparatus |
-
1990
- 1990-09-25 US US07/587,913 patent/US5191636A/en not_active Expired - Lifetime
-
1991
- 1991-09-23 EP EP91916369A patent/EP0550501B1/en not_active Expired - Lifetime
- 1991-09-23 DE DE69105145T patent/DE69105145T2/en not_active Expired - Fee Related
- 1991-09-23 CA CA002091772A patent/CA2091772C/en not_active Expired - Fee Related
- 1991-09-23 WO PCT/CA1991/000338 patent/WO1992005478A1/en active IP Right Grant
- 1991-09-23 JP JP3515417A patent/JPH06501577A/en active Pending
- 1991-09-23 AU AU85222/91A patent/AU8522291A/en not_active Abandoned
- 1991-09-23 ES ES91916369T patent/ES2063524T3/en not_active Expired - Lifetime
Non-Patent Citations (2)
Title |
---|
CENTRAL PATENTS INDEX , BASIC ABSTRACTS JOURNAL Section T, week 9023 * |
SCIENTIFIC HONEYWELLER. vol. 10, no. 1, 1989, MINNEAPOLIS US pages 67 - 79; A. GUHA, A. HAGGERTY, J. JELINEK: 'Neural Networks as a Control Technology' see page 68, left column see page 71, left column, line 21 - line 47 * |
Also Published As
Publication number | Publication date |
---|---|
DE69105145T2 (en) | 1995-04-06 |
DE69105145D1 (en) | 1994-12-15 |
US5191636A (en) | 1993-03-02 |
CA2091772A1 (en) | 1992-03-26 |
CA2091772C (en) | 2001-01-02 |
EP0550501A1 (en) | 1993-07-14 |
EP0550501B1 (en) | 1994-11-09 |
AU8522291A (en) | 1992-04-15 |
ES2063524T3 (en) | 1995-01-01 |
JPH06501577A (en) | 1994-02-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US5191636A (en) | Control system with two level sensor system | |
Rummery et al. | On-line Q-learning using connectionist systems | |
US8856057B2 (en) | Cognitive security system and method | |
Webb | Robotic experiments in cricket phonotaxis | |
CN109114411B (en) | Access sensing system for security incidents | |
EP0865989A3 (en) | Seal integrity monitoring and adaptive control method and apparatus | |
CN107092259A (en) | The anti-fall method and device of robot | |
CN111721057A (en) | Control method, control device, refrigeration equipment and storage medium | |
WO2008033118A1 (en) | Door assembly including a sensor for controlling automated door movement | |
CN110812740A (en) | Control method of fire extinguishing bomb in tunnel and related device | |
Hirel et al. | Biologically inspired neural networks for spatio-temporal planning in robotic navigation tasks | |
CN105528892B (en) | A kind of method and device for solving the traffic congestion of parking exit passage | |
Timmis et al. | A neural-endocrine architecture for foraging in swarm robotic systems | |
CN114291672A (en) | Elevator riding control method and system for robot, robot and storage medium | |
Christensen et al. | Automatic synthesis of fault detection modules for mobile robots | |
CN108082603B (en) | Method and system for automatically controlling material processing | |
Watson et al. | Optimal decision making in robotic assembly and other trial-and-error tasks | |
Gadanho et al. | Learning behavior-selection in a multi-goal robot task | |
CN109625814A (en) | The control system and method for transport vehicle | |
CN114394105B (en) | Intelligent driving system management method | |
Nassery et al. | Real time seismic signal processing using the arma model coefficients and an intelligent monitoring system | |
Ni | 3.3 Algorithm Performance | |
JP3082806B2 (en) | Fault diagnosis device | |
Milighetti et al. | Multi-sensor controlled skills for humanoid robots | |
JPH0431554B2 (en) |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AU BB BG BR CA CS FI HU JP KP KR LK MC MG MN MW NO PL RO SD SU |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): AT BE BF BJ CF CG CH CI CM DE DK ES FR GA GB GN GR IT LU ML MR NL SE SN TD TG |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2091772 Country of ref document: CA |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1991916369 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 1991916369 Country of ref document: EP |
|
WWG | Wipo information: grant in national office |
Ref document number: 1991916369 Country of ref document: EP |